Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old April 8th 04, 11:10 PM
Robert Richter
 
Posts: n/a
Default Offering alternate Nalimov files with pawns

I posted just a few minutes ago (but it doesn't show up yet, so
starting a new thread). With incomplete tablebases and pawns, the
computer won't play right because it refuses pawn promotion. How
about making search engines capable of an alternate format (about 16%
larger???) that includes the pawn on the 8'th row as queens. This
way, if you only download the tablebases with pawns, it will see that
queening is the right thing to do, and queen the pawn, even if that
throws the game back into the search engine; as it is, the chess
engine will not work if queening puts it into a tablebase that is not
included on your HDD. If queening is the wrong thing to do and
underpromotion is better, well, you will need tablebases to detect
that, but this slightly larger file is certainly better than the
following:

kppkpp won't work unless I have
kqpkqp, kqqkqp, kqqkpp, kqqkqq, which is four 6 man tables I would
have to download, and I may only have room for one 6 man table, so
make it 16% larger so these other files don't have to be downloaded,
as an alternate format, of course, so as to not ruin what has already
been done. Perhaps a simple took can be written to merge the queen
and pawn files so that tablebases don't need to be regenerated.

The situation I ran into with 5 man is not a problem for a lot of
people. A lot of people have hard drives capable of handling the
entire 5 man set so it will work properly, but with the 6 man
tablebases being generated, this may be a problem.

Sorry if I am just a beginner and this has already been discussed; I'm
not a skilled chess player and don't follow these forums closely.

PS--16% comes from assuming pawns on 2-7, allowing pawns on 2-8 (but
queens when on 8) allows 2-8, which is 16% larger than 2-7. Is this
correct thinking?
  #2   Report Post  
Old April 9th 04, 01:58 AM
Robert Hyatt
 
Posts: n/a
Default Offering alternate Nalimov files with pawns

Robert Richter wrote:
I posted just a few minutes ago (but it doesn't show up yet, so
starting a new thread). With incomplete tablebases and pawns, the
computer won't play right because it refuses pawn promotion. How
about making search engines capable of an alternate format (about 16%
larger???) that includes the pawn on the 8'th row as queens. This
way, if you only download the tablebases with pawns, it will see that
queening is the right thing to do, and queen the pawn, even if that
throws the game back into the search engine; as it is, the chess
engine will not work if queening puts it into a tablebase that is not
included on your HDD. If queening is the wrong thing to do and
underpromotion is better, well, you will need tablebases to detect
that, but this slightly larger file is certainly better than the
following:


This just won't work. The tables don't have pieces in them. They are indexed
by the position of the pieces, and krpkr would not have any "queen" info in it
at all...


kppkpp won't work unless I have
kqpkqp, kqqkqp, kqqkpp, kqqkqq, which is four 6 man tables I would
have to download, and I may only have room for one 6 man table, so
make it 16% larger so these other files don't have to be downloaded,
as an alternate format, of course, so as to not ruin what has already
been done. Perhaps a simple took can be written to merge the queen
and pawn files so that tablebases don't need to be regenerated.


The situation I ran into with 5 man is not a problem for a lot of
people. A lot of people have hard drives capable of handling the
entire 5 man set so it will work properly, but with the 6 man
tablebases being generated, this may be a problem.


Sorry if I am just a beginner and this has already been discussed; I'm
not a skilled chess player and don't follow these forums closely.


PS--16% comes from assuming pawns on 2-7, allowing pawns on 2-8 (but
queens when on 8) allows 2-8, which is 16% larger than 2-7. Is this
correct thinking?


--
Robert M. Hyatt, Ph.D. Computer and Information Sciences
University of Alabama at Birmingham
(205) 934-2213 136A Campbell Hall
(205) 934-5473 FAX Birmingham, AL 35294-1170
  #3   Report Post  
Old April 9th 04, 08:39 AM
Anders Thulin
 
Posts: n/a
Default Offering alternate Nalimov files with pawns

Robert Richter wrote:

computer won't play right because it refuses pawn promotion. How
about making search engines capable of an alternate format (about 16%
larger???) that includes the pawn on the 8'th row as queens.


That would mean that the chess engine would need to be modified, and
there would also need to be some kind of reconstruction of pawn-ful position
to include those extra positions. And some kind of code to handle the
normal queen-less files as well as the new pawn-and-queen games.

But in that case, why not modify the chess engine to be more robust
when tablebase files happen to be missing?

way, if you only download the tablebases with pawns, it will see that
queening is the right thing to do, and queen the pawn,


Much more efficient to encode that rule (in case of a missing
tablebase after promotion, always queen) in the chess engine.
Depending on the engine, it might actually be better to ignore
the tablebases, and rescore that position using the usual scoring
function instead.

Anyay, modifying the engine will require much less space than
8% of each pawn-ful endgame. It'll probably be done much
faster as well ...

--
Anders Thulin ath*algonet.se http://www.algonet.se/~ath
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT +1. The time now is 10:55 AM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017